4/6/2022

Dissertation Outline

Theme: Influence of human activity on landscapes and predator-prey interactions

  • Ch. 1: Rapid islandization of Earth’s protected areas (In Prep)
  • Ch. 2: Habitat edges drive animal movement in protected areas (TBD)
  • Ch. 3: The influence of human activity on predator-prey interactions (In Review)
  • Ch. 4: Characterizing coyote resource selection in human-modified landscapes (Analysis)

Research Objectives

Characterize the drivers of coyote behavior and resource selection at an active sheep ranch.

Data Collection

Dataset:

  • 2 field seasons, December - March
  • 8 coyotes, 1-hour fix rate
  • 4 guardian dogs, 10-min fix rate

Notes:

  • Lambs in pasture February - March
  • Grapes in season August - November

Coyote Home Ranges

These are the 95% Kernel Utilization Distributions for 2 seasons of coyotes (n = 9).

Covariates

There are several environmental variables that may influence coyote resource selection, including:

Physical Other
Land use Day/Night
Slope Lamb presence
Elevation Guard Dog home range
Distance to water Grape presence

Covariates

Step Selection Functions (SSF)

A first pass: Land Use

How do coyotes use cropland and natural areas?

Steps inspired by Abrahms et al. 2016:

  1. Use HMM to parse into three behavioral states (resting, traveling, meandering).
  2. Create a SSF for all movement data without parsing by behavior (‘combined model’)
  3. Create SSFs for movement by behavior (‘behavior model’)
  4. Compare

Hidden Markov Models (HMM)

Hidden markov models can use distance traveled and angles between time points to infer behavioral states of animals.

Hidden Markov Models (HMM)

Initial parameters ranges for 3 states:

State SL_min SL_max SD_min SD_max TA Conc_min Conc_max
Resting 50 100 25 50 3.141593 0.2 0.5
Meandering 500 1000 250 500 1.570796 0.5 0.7
Traveling 1000 3000 500 1500 0.000000 0.7 3.0

Hidden Markov Models (HMM)

Here are the parameters for the best HMM (n = 25 iterations), with the lowest AIC.

## [1] "Step Parameters:"
##                state 1      state 2      state 3
## mean      2.582279e+01 1.869097e+02 7.724008e+02
## sd        1.963225e+01 1.589263e+02 5.841408e+02
## zero-mass 4.166166e-04 1.010215e-03 2.473974e-04
## [1] "Angle Parameters:"
##                 state 1   state 2     state 3
## mean          3.0936337 3.0003092 -0.05766519
## concentration 0.4500531 0.3069862  0.33858379

Hidden Markov Models (HMM)

## Decoding states sequence... DONE

Hidden Markov Models (HMM)

We can append the inferred state to the dataframe:

## # A tibble: 6 × 7
##   ID     step  angle       x        y time                states
##   <chr> <dbl>  <dbl>   <dbl>    <dbl> <dttm>               <dbl>
## 1 C4    358.  NA     494641. 4319299. 2021-02-28 17:00:15      2
## 2 C4    388.   0.289 494510. 4318966. 2021-02-28 16:00:16      2
## 3 C4    332.  -1.63  494477. 4318580. 2021-02-28 15:00:12      2
## 4 C4    141.   2.36  494148. 4318628. 2021-02-28 14:00:30      2
## 5 C4     34.2 -1.51  494233. 4318514. 2021-02-28 13:00:19      1
## 6 C4     20.9 -0.601 494207. 4318492. 2021-02-28 12:00:19      1

Step Selection Functions (SSF)

Model structure: amt::fit_issf(steps, case_ ~ landuse_end + strata(step_id_))

Table 1.Summary of step selection coefficients for ‘Land Use’ by behavior category (n = 7 individuals)

Behavioral State Observed Steps Estimate (Cropland) SE Estimate (Natural) SE2
Resting 4946 8.060761 995.5901 5.588340 751.0802
Meandering 5000 7.395303 1480.1635 4.275815 845.9252
Traveling 9465 7.637681 1326.9548 6.532572 1137.4405
Combined 12823 6.717207 786.1172 6.834528 786.1084

**Note: These standard errors are wild, and the estimates also look strange…

Initial conclusions

Figure 1. Coyote selection for croplands and natural areas near Hopland, CA (n = 7). * indicates estimate means.

Next Steps

Measure influence of:

  • Guardian dogs on coyote behavior
  • Lambing period on coyote resource selection
  • Day / Night on habitat use

Consider:

  • Model selection framework to test other covariates
  • 3rd field season (is timing important? number of individuals?)
  • Other ideas?